Assuming that you want to use a simple raster grid solver, surface currents can (to a first approximation) be calculated as the gradient of a distance field calculated from sea portion of the altitude field after a bit of blob-finding for large, deep-ocean areas. Clockwise gradient above the equator, counter-clockwise below. I recommend using an equal-area map projection for the calculations, but connectivity is a tough one for most projections. Note that I say "to a first approximation", because it doesn't account for things like splitting or merging currents. You might be able to feed the simple vector field above into something like a stable fluid simulator, but I suspect that it won't be stable long.
Air currents are approximately the same, but you need to take into account the ITCZ and summer highs vs. winter lows. For moving heat, you need to have a sea<->air and land<->air interchange rates. For moving moisture (rain), the air parcels need to charge over water at a rate dependent on temperature and humidity, and discharge over land at a rate dependent on biome (humidity) and tempearture (including drops due to altitude).
If you're using something like a Delaunay graph through points distributed worldwide, then I'm not sure that I can offer much in the way of suggestions except distance from shore and gradient of that distance. It's the same idea as above, but I'm not sure how to go about solving the distances easily.
All of the heuristics aren't needed with detailed fluid sims drive by asymmetric solar heating and a spinning globe, but that's been pretty ugly historically. It might be tractable (at least coarsely) on some of the larger variants of today's systems. I've found that the choice of data representation is the hard part.